Six-week-old female C57BL/6 mice [Animal BioResource (NSW, Australia) or Animal Resources Centre (WA, Australia)] were housed under specific pathogen-free conditions in the animal facility of the Charles Perkins Centre. Diets were purchased from Specialty Feeds (Glenn Forest, Australia) and mice were fed for 2 weeks prior to infection, either with commercially available diets deficient in dietary fiber (SF11-028), or enriched in dietary fiber (SF11-025: resistant starch gel crisp as source of fiber) as previously described (3). Experiments were performed in accordance with the animal ethics protocol 2016/976 approved by The University of Sydney Animal Ethics Committee.

The original stock of WNV (lineage II Sarafend strain), acquired from The John Curtin School of Medical Research (ACT, Australia) was propagated alternately in C57BL/6 suckling mouse brains and in vitro in Vero cells (24). Mice anesthetized intraperitoneally with Avertin at day 0 were inoculated intranasally with 6 × 10⁴ and 6 × 10³ plaque-forming units (PFUs) of WNV to achieve 100% lethal dose (LD₁₀₀) and 50% lethal dose (LD₅₀) studies, respectively, as described in (27). Alternatively, mice were injected with a LD₁₀₀ intracranially via the postglenoid foramen, an approach that does not penetrate the skull bones, minimizing tissue damage (28). Mice were weighed daily, and assessed for clinical symptoms, as previously described (24). Briefly, they were scored as follows: score 0: no clinical signs, 1: Weight loss < 5%, 2: Weight loss ≥ 5%, 3: Weight loss ≥ 5% with significant reduction in movement, 4: Weight loss ≥ 5% with significant reduction in movement, ruffled fur, hunched posture, and seizures and 5: Immobile, cold.

Mice were anesthetized and perfused with ice-cold PBS before collection and processing of tissues into single cell suspension as previously described (26, 29). Dead cells were excluded, based on their staining with the LIVE/DEAD™ Fixable Blue Dead Cell Stain (ThermoFisher Scientific) and anti-mouse CD16/32 (BioLegend) was used to prevent antibody non-specific binding. Cells were permeabilized with the Cytofix/Cytoperm kit (BD Biosciences) for intracellular staining. Antibodies used for flow cytometry are listed in Table S1 . To study cytokine expression, cells were stimulated with 50ng/ml phorbol 12-myristate 13-acetate, 500ng/ml ionomycin and 5µg/ml brefeldin A for 4 hours. Flow cytometry was run on the LSR-II analyser (Becton Dickinson, San Jose, CA, USA) using the FACSDiva software and data was analyzed with FlowJo v10.7.1. (Treestar Inc. Ashland, OR, USA) based on gating strategies presented in Figures S2-3 .

Total tissue RNA was extracted using TRI Reagent (Sigma Aldrich), based on the manufacturer’s protocol. cDNA was synthesized with the High-Capacity cDNA Reverse Transcription Kit (ThermoFisher Scientific). qPCR was conducted with the Power SYBR™ Green PCR Master Mix (ThermoFisher Scientific) on LightCycle^® 480 Instrument II (Roche) with primers listed in Table S2 .

Paraformaldehyde-fixed (4% in PBS), paraffin-embedded colon tissues were sectioned and stained with H&E and imaged using light microscopy (Zeiss Axioscope, Zeiss, Oberkochen, Germany). Colonic inflammation was scored, based on previously published guidelines (30).

For immunohistochemistry, colon tissue was fixed overnight in 2% PFA and subsequently placed in a series of solutions of progressively increasing sucrose concentration (10%, 20% and 30% sucrose in PBS), before being embedded in optimum cutting temperature compound (O.C.T.; Tissue-Tek) and frozen in hexane pre-chilled in liquid nitrogen, as previously described (26, 31). Tissue blocks were sectioned (8–9 µm), fixed in methanol, rinsed in tris-buffered saline with 0.05% Tween 20 (TBST) and blocked with 10% FCS before being stained with primary fluorophore-conjugated antibodies targeting WNV non-structural protein 1 (NS-1) and FOX-3 in neuronal nuclei (NeuN). Tissue sections were washed twice in TBST before being counterstained with DAPI antifade (Vector Laboratories). Images were acquired on the Olympus BX51 Microscope using a DP-70 camera and Cell Sensor software.

Fecal SCFA acetate was quantified by ¹H nuclear magnetic resonance spectroscopy as previously described (11). Feces were first homogenized with deuterium oxide (Sigma-Aldrich) at a concentration of 100mg/ml and centrifuged for 5 minutes at 14000 × g at 4°C. The resulting supernatant was then diluted in sodium triphosphate buffer (pH=7.0) (Sigma-Aldrich), with 0.5mM 4,4-dimethyl-4-silapentane-1-sulfonic acid as the internal standard (Sigma-Aldrich). Samples were run on Bruker 600 MHz AVANCE III Spectrometer and analyzed with the Chenomx NMR Suite v8.4 (Chenomx Inc.).

DNA from fecal samples were extracted using the FastDNA Spin Kit for Feces (MP Biomedicals) following the manufacturer’s protocol. The primers for 16S microbiota analysis were selected to amplify the V3-V4 region of 16S rRNA genes. Forward primer used was ACTCCTACGGGAGGCAGCAG and reverse GGACTACHVGGGTWTCTAAT. The primers contained barcodes, spacers and Illumina sequencing linkers and were designed and used as suggested by Fardosh et al. (32). The 16S rRNA gene sequencing library preparation, PCR amplification and library purification followed the Illumina recommended protocol (Illumina Inc., San Diego, CA, USA). The sequencing was performed on an Illumina MiSeq instrument using 2x300 bp paired-end sequencing. Amplicon sequence variant was generated with the dada2 package (1.16.0) using R software (4.0.2). Taxonomy was assigned using the Ribosomal Database Project classifier with species level taxonomy assignment. Alpha and beta diversity analysis was performed using the phyloseq (1.32.0), microbiome (1.10.0) and vegan (2.5-7) packages. Sequence data was deposited in the European Nucleotide Archive under accession number PRJEB50194.

Unpaired t-test and ANOVA were used to compare two or more groups, respectively, and Mantel-Cox log-rank test was used to analyze LD₅₀ study survival results. Differences were considered statistically significant when p<0.05.

To determine whether beneficial reshaping of the gut microbiota could affect the severity of WNV infection, mice were fed on diets enriched in dietary fiber (high fiber, HF) or deficient in dietary fiber (zero fiber, ZF) for 2 weeks prior to infection ( Figure 1A ). These diets have previously been shown to reshape the gut microbiota composition beneficially (HF) or detrimentally (ZF) (3, 4). Mice fed on HF, as expected, had a significantly different gut microbiota composition as shown by weighted UniFrac ( Figure S1 ), as well as a significantly increased production of the SCFA, acetate, in feces, compared to ZF-fed mice ( Figure 1B ), as previously published (5). These mice were inoculated intranasally with WNV LD₁₀₀ and maintained on these diets over the course of infection. Intranasal inoculation enables reliable infection of the central nervous system (CNS), with clear separation of anti-viral responses in the CNS from systemic responses. In this model, WNV spreads from the olfactory bulb to the rest of the brain and spinal cord over the course of infection and has been described in detail by Getts et al. (26).

Both HF- and ZF-fed mice had altered gut microbiota composition after WNV infection ( Figures S1A–G ), with the bacteria from the genus Dorea, Enterorhabdus and Clostidium IV being significantly more represented in infected ZF-fed mice. Despite these differences, acetate levels remained higher in infected HF-fed mice ( Figure S1H ). Nevertheless, both groups had similar clinical scores at 7 days post infection (dpi) ( Figure 1C ), as well as undergoing comparable weight loss ( Figure 1D ). This suggests that HF diet does not protect against LD₁₀₀ WNV infection. Consistent with this, total numbers of leukocytes in the brain and infiltration of inflammatory Ly6C^hi monocytes was similar in both groups, as assessed by flow cytometry, although neutrophil infiltration was significantly reduced in HF-fed mice ( Figure 1E and Figures S2, S3 ). The aggravating effects of antibiotics have been linked to impaired WNV induced-T cell responses (22). In contrast, HF and SCFA can modulate T cell differentiation towards both Th1, Th17 and Treg, depending on the context (12). However, numbers of infiltrating CD4⁺ and CD8⁺ T cells into the brain were similar between HF and ZF groups ( Figure 1F ). Furthermore, in WNV infection, HF feeding did not affect the T cell response, with similar numbers of Th1, Th17 and Treg, as well as IFN-γ⁺ CD8⁺ T cells, both in the brain ( Figure 1F ) and its draining lymph nodes, when compared to ZF ( Figure S4 ).

As beneficial immunomodulation by HF diet may be masked by overwhelming infection at the LD₁₀₀, we investigated the effects of dietary fiber in WNV infection at LD₅₀ and assessed its impact on survival. In mice inoculated intranasally with LD₅₀ as in Figure 1A , clinical disease scores increased to a similar extent in both dietary interventions from 5 dpi onwards ( Figure 1H ). By 10 dpi, independent of diet, the mortality of both groups was similar at 30-50% and remained stable after 16 dpi ( Figure 1G ). We confirmed that the absence of protection in the HF group was not linked to this diet in particular, as a different high-fiber diet enriched in guar gum, previously shown to improve health (3–5), also had no impact (data not shown). Taken together, these data indicate that dietary fiber does not reduce the clinical severity of WNV encephalitis.

Systemic infection associated with peripheral WNV inoculation has been linked to enteric neuronal infection and neuronal injury promoted by antigen-specific CD8⁺ T cells (33), resulting in altered gut homeostasis, in particular, gastrointestinal motility (33). Whether CNS infection with WNV is also associated with WNV in the gastrointestinal tract is unknown. To address this question, we inoculated mice fed on normal chow diet intranasally (i.n.) with LD₁₀₀ WNV and assessed the presence of WNV by qPCR in the brain and brain-draining cervical lymph nodes (cLN), the gastrointestinal tract (duodenum, jejunum, ileum and colon), the gut-draining mesenteric lymph nodes (mLN), the primary lymphoid organs (thymus and spleen) and peripheral organs (lung, kidney and heart). While the brain was the major site of infection, we also detected significant viral load in the colon and at lower levels in the duodenum, jejunum, ileum and their mLN, as well as the heart. Viral load in the colon was even higher in animals inoculated intracranially (i.c.) with WNV, excluding the possibility that infection was a result of systemic infection or ingestion ( Figure 2A ). We confirmed the presence of WNV by immunofluorescence in the colon, specifically in colonic neurons of the myenteric and submucosal plexuses, colocalizing with the neural marker NeuN ( Figure 2B ). We then assessed whether the presence of WNV was associated with colonic mucosal inflammation. Histologically, mice infected with WNV had significantly more mononuclear cells infiltrating the colon ( Figure 2C ), and this was associated with a significantly reduced colon length in infected mice ( Figure 2D ), as well as changes in the gut microbiota composition ( Figure S5 ). Thus, CNS infection with WNV is accompanied by spread of virus in the gastrointestinal tract, particularly in colonic neurons, leading to local inflammation.

As dietary fiber is known to promote gut health, we then investigated whether HF affected viral load in the colon. Regardless of diet, mice fed on HF or ZF had similar colonic viral load ( Figure 3A ), showing no effect of dietary fiber. The presence of virus was associated with significantly increased mRNA expression for type 1 and type 2 IFNs, IL-6, tumor necrosis factor (TNF) and IL-10 expression in the colon in both groups ( Figures 3B–F ). While levels of mRNA for these cytokines were similar between groups, mice fed on HF had significantly decreased levels for TNF, compared to mice fed on ZF, suggesting potential local immunomodulatory effects of HF in the colon ( Figure 3E ). The similar overall levels of message for cytokines (except for TNF) correlated with similar increased mononuclear cell infiltration by histological analysis between groups (data not shown). Infected HF-fed mice also had immune phenotypic profiles in the mLN similar to infected ZF-fed mice ( Figure S6 ), suggesting a similar response to infection in both groups.

The fermentation of dietary fiber affects both the colon and cecum length (3). Accordingly, the colons of mock-infected HF-fed mice were significantly longer than ZF-fed mice. However, in WNV infection by 7 dpi, colon length in HF-fed mice was significantly reduced and comparable in size to those of ZF-fed mice. Notably, the colon length in infected ZF-fed mice was similar to mock-infected ZF-fed mice ( Figure 3G ).

Together, these data show that WNV infection of the CNS is accompanied by infection of the autonomic nervous plexuses in the colon, mucosal mononuclear cellular infiltration, local elevation of immune cytokines and reduced colonic size, consistent with inflammation, which, despite reduced TNF levels, was unaffected by the fiber content in HF feeding.